Tag Archives: good-making features

Theory, empirical data, appearances and illusion


Time and the metaphysical condition of illusion

What illusions do theories of perception have to explain?


Of all my research, this is the area of which I am least sure. It seems to me that much of the philosophy of science literature concerned with metaphysics (and not, e.g., epistemology) is about questions such as, e.g.,

(a) How necessary is realism about scientific entities for theory — e.g., must electrons be real for them to be useful for explanation?


(b) Is empirical evidence a sufficient or exhaustive means of evaluating competing scientific theories?

I think what I’m interested in here is related to these questions (as seen below) but it is not the same. Instead, my concern is with how one uses empirical data in evaluating scientific theories.

  • Unlike ‘a’, I am neutral in regards to the reality of entities posited by such theories.
  • Unlike ‘b’, empirical data is a means of evaluating two competing theories; it concerns cases of evaluation falling within empirical evidence. Although it does not require that empirical data is the only kind of evaluation (what I take to be the issue with ‘b’).

What I express here might be plausible, necessary, contingent, or impossible. In any case, I am expressing it from a position of unreflective — or barely reflective — intuition; what I say here seems prima facie plausible to me. In the context of this project, I think it is fair to start with this, and modify it as I go. But I cannot pretend that this is not a naïve point in my research.

In any case, I’ll press ahead.

My first thought about empirical data and scientific theory is this:

  • Correspondence of a scientific or empirical theory to empirical data is a good-making feature of that theory.  Empirical data is a way of evaluating competing scientific or empirical theories. If some empirical datum = ‘x in situation E’, then if: ET1 (a scientific or empirical theory that predicts or claims ‘x in situation E’) and ET2 (another scientific or empirical theory that does NOT predict or claim ‘x in situation E’, or even DENIES ‘x in situation E’), then ET1 is a better empirical or scientific theory than ET2.
    • Some theories cannot be compared this way. They both predict or claim ‘x in situation E’ and are both consistent with that empirical datum, or both do not predict or claim, or both even deny, the datum. As such, this empirical datum is not a way of evaluating these theories. Such a situation, I am given to understand, is described under the Dukem-Quinethesis (in Stanford 2009).
      • First, this neutrality regarding empirical data is not the situation I am interested in; I am concerned by situations where, prima facie at least, one theory predicts or accepts the datum, the other does not predict it and/or denies it.
      • Second, I am not sure that a point debated between two theories which is not distinguished by an empirical datum is, relative to that empirical datum, rightly thought of as a point debated by these theories as scientific or empirical theories. What would make them scientific or empirical theories is something else, some other way in which they are tested by empirical data. Leading to this:
        • If one were to have two theories which were both consistent with all of the same empirical data, then what distinguishes them, and thus what is only relevant to evaluating one over the other, does not fall within the domain of science or empiricism. It is, in my view, then a metaphysical issue.

    My worry about illusions(and hallucinations) is that any casual use of illusions and hallucinations to remove appearances allows one to neutralise the evaluative force of empirical data. In doing so, by removing the force of appearances, it turns claimed scientific or empirical disputes into metaphysical disputes.

    • By ‘casual’ I mean one claims, for whatever reason, that ‘x is an illusion’ and then leaves it at that, without explaining hownot why — such an illusion occurs.
    • For a scientific theory, casual positing of illusions and hallucinations from within one’s theory is a bad-making feature of that theory as a scientific or empirical theory.
      • But it’s grand as a metaphysical theory. Metaphysical theory is entitled to use whatever works to explain the world, so long, I think, as it is coherent.
        • If you want to claim that everyone everywhere is constantly in perceptual error, except on precisely the winter solstice, then if that is coherent, and explains the world better than alternatives, go for it. Just don’t pretend that your theory has not become a metaphysical theory.
      • By ‘from within one’s theory’ I mean that the illusions and hallucinations are not already there in one’s domain of enquiry. If the illusions and hallucinations are not from within one’s theory, they are ‘from outside one’s theory’. In that case,  they would be forced upon us with or without that scientific/empirical theory. In that case, they are not a bad-making feature of that theory.
        • Or, I should say, they are not a bad-making feature of only that theory. It may be that many competing theories all involve commitment to these illusions and hallucinations, and so cannot be evaluated by them in comparison to each other. But there may be another theory, or group of theories, which are not committed to these illusions and hallucinations. If so, regarding these illusions, the first group are negatively evaluated in comparison to the second group — again, as scientific and empirical theories (as metaphysical theories, there is no problem).
  • A central component of empirical data is what is apparent. I think this is an intuitive point, but as will be discussed it might be controversial for some. When something is empirical, when it is empirical data, it is constrained as follows (in my view; again, this may very well look like a very naive view).
    • Empirical data needs to happen in time. The observation is not of Platonic truths such as those of mathematics. Reason, or the rationality, is generally held to give us access to those, and reason has historically been considered an alternative to the empirical. (This is not to say that it is not itself a kind of perception — Godel, for one, thought that we perceive mathematical truths similar to how we perceive distant mountains (Yourgrau 2005))
    • Empirical data also has to involve appearances in some way. I would say that it is not something we only infer from appearances, but includes appearance themselves. How things seem is part of the empirical data. One could also include, and in practice one does, what it is plausible to infer from appearances.
    • Empirical data is public: a shared entity to which separate individuals have access.
      • I do not think this is necessary for a scientific or empirical data. Hume may be the only person alive, and the only person with access to data, but he may still be entitled to call it empirical data. However, considered in the context of its role in a scientific or empirical theory, there are other conditions of such a theory in the actual world: such a theory is a shared theory, with data which different individuals can show to each other. That is, the data, the empirical data, is public in actual world theories.
      • This raises the possibility of the following tension: What is apparent is part of empirical data. But empirical data is also public. Yet, one might object to this for the following reasons: (a) what is apparent is not public; it is particular or private to each individual or (b) assuming that what is apparent is public to assume to assume a disputed metaphysical position: naive or direct realism, perhaps.
        • In particular, this tension comes where what one is referring to in referring to appearances is to what some others may call ‘phenomenal’ or ‘qualia’. The what it is like of experience is apparent to one but is not public, so this thinking might go.
          • In response, one might say: qualia shows that some appearances are private. In that case, they are not empirical data. But those appearances that are public can be empirical data.
          • Alternatively, I suppose someone might claim they are public enough to be empirical data because each subject understands what other subjects mean by such qualia, e.g., the red-ness or blurriness, etc., although they do not have access to the specific case of the other subject’s qualia. Not only is there something it is like to see red, but lots of people know what it is like, and so it is something we can discuss publicly. But my thinking is that ‘public’ in this sense refers to be reports by subjects, not the qualia themselves (and so we have the sort of distinction made by, e.g., Dennett’s heterophenomenology).
        • But one might still object if one held that all appearances are private and not public. This is especially the case for those who hold sense-data theory. Sense-data are the bearers of properties that external things seem to bear, e.g., colours, shapes, and so on. Sense-data (like qualia) are private, not public, and so, by the definition suggested here, cannot be empirical data. But if they are not empirical data, what is?
          • I take it that such problems are problems — this is a tension for empiricists who hold to appearances being private (as, e.g., sense-data). I think sense-data theorists should offer an explanation for it. But still: if they deny that appearances play a role in empirical data, I think that this is also a problem. I leave it to advocates of such views to sort it out.
          • I suppose one way around it is to note that, by motivation of their introduction to theory, such entities as sense-data entail a public world which, in most cases, causes the sense-data; I’m thinking here of the sort of process from sun shine to sense-data one gets in writers such as Russell [reference pending, but check my 2010b for brief discussion concerning time-lag]. Then, in cases where they happen all on their own, these are illusions and/or hallucinations., and are not empirical data.
    • Appearances, and what we infer from them, play the evaluative role in scientific and empirical theory. If a scientific or empirical theory matches the appearances, this is good; if it does not, this is bad.
      • This — I take it — is what is meant by ‘saving the appearances’.
      • The ‘what we can infer’ shows one way out with illusions and hallucinations.
        • One rejects the appearances and does not save them because they are appearances in cases of illusions or hallucinations.
        • However, as discussed above, I consider doing this to be a bad move for a scientific or empirical theory.
        • One may be forced into that move for a scientific or empirical theory. But if your competitor is not forced into it, then they have the advantage.

Conclusion of this section

I’ll leave this here for now. As I said, this is primarily a laying out, with some basic justification, of my intuitions about empiricism, appearances, scientific theory and illusion. I expect to modify it from any comments I get, discussions, or publications.

The main conclusion I think is this:

1. Appearances, as components of empirical data, evaluate the claims of scientific and empirical theories.

2. If you dispense with appearances in your theory, then your theory is not a scientific or empirical theory.

3. If you casually assert illusions or hallucinations as an explanation of appearances, then you dispense with the appearances.

4. So, if you casually assert illusions or hallucinations as an explanation of appearances in your theory, then your theory is not a scientific or empirical theory.

‘4’ is a very strong claim. I am aware of that, but do not know what an alternative position to take about scientific/empirical theory and illusion/hallucination could be. In science/empiricism, appearance matters to ontological claims; in illusion/hallucination, appearance does not matter to ontological claims.

Still: perhaps we could remove appearances, remove how things seem, and try to re-define a notion of empiricism without it. Given the angle of my work here, I can see why this would be a desideratum of theorists who want to remain proposers of scientific theories but want to ignore appearances.

To my mind, this revisionary conception of empiricism is not ideal. I do not know how one can say that there is empirical evidence of some x if at some stage there is not an appearance of x or an appearance from which one might infer x. That is, although scientific (or other empirical) claims might go beyond the appearances and naive assertions about reality, I take it that, as claims that are scientific (or empirical), they are grounded and constrained by how things seem.

Even if experimental design involves unobservable events, even if one’s understanding of the resulting data is exhaustively contained within a language or theory with necessary and prior metaphysical commitments (as ‘theory-laden’ debates concern, e.g., ): if it is scientific or an empirical theory, and not a , I take it appearances are part of the evidence. They are part of how one confirms or falsifies the theory; they increase or decrease the probability of the theory being true (e.g., Williamson 2000).

But this is all by the wayside if it is only a naïve view, if this is only how it seems to me given my current understanding. If presented with sound or at least valid arguments, perhaps I can be persuaded away from this thinking.


What illusions do theories of perception have to explain?

Back: Principle of Universal Illusory Counterpart

[You can see this post as connecting to and overlapping a little with my 2018 book‘s chapters on different kinds of experiential error and choosing between them. But this is about something else, whether or not we should explain merely possible illusions or not (I think not) and what that means for theories of perception.]


I show you an image and ask you ‘what do you see?’

You answer that you see, e.g., ‘such-and-such x’, some kind K, property P, relation R, or a combination of similar things — particulars, kinds, properties, relations.

And I answer: ‘ok. But there are no K, P, R, x, or combination of these or things similar. This is an illusion. ‘
In response, you say ‘wow…’ (Because you would) and then ‘but how can there be this illusion?’

What should I say?

  • I might say ‘because how you see can lead to illusions and this is one of them.’ However, this is not an answer involving new information. If I’m assuming that the illusion comes from my seeing, I already know that seeing leads to illusions, and this is one of them.
  • So, I might explain how perception works. In so doing, I explain how the way perception works can sometimes result in the illusion I demonstrate here. That is, my theory of perception explains this instance of illusion.[1]


Generally, what cases of illusion should I be explaining here? Need I only explain illusions that are actual or must I also explain illusions that are merely possible? In this section, I discuss why I think we need only explain actual illusions. This is not to say we cannot explain merely possible illusions, but that our theories ought only to explain the actual ones. (The section to follow, on what actual illusions show, discusses what we can predict from such actual illusions as regards possible illusions).

Universal Illusion

Like Uncle Petros dreaming the solution to Goldbach’s Conjecture, what might be apparent could extend very far, and fail to correspond to how things are. As said before, some might use ‘appearance’ to correspond to what one believes, and this can include mathematical claims, and beliefs about abstract entities such as universals, unobtained possible world, etc. There are also those who think that some ways of understanding such abstract entities are less like some general belief and much closer to the perception of the world, e.g., Godel thought his understanding of maths was a form of perception, a perception of Platonic mathematical entities (Yourgrau 2005)).

Here, I want to narrow the appearance of a mind-independent world, and that it is only a dream, to concrete particulars and their exemplification of properties, relations, etc. With that, I think we can say the following:

  • Seemingly, I perceive and remember particular spatially, temporally located entities and the properties that they possess. Consistent with such an appearance, or such phenomenology, is that (for perception) there are no and (for memory) there never were any of these particular spatially, temporally located entities and properties.
  • If there are or never were such spatially, temporally located entities, my entire experience is an hallucination.
  • If there are (and were) particular entities, but they have none of the properties that I seem to perceive, my entire experience is an illusion.

Illusions will only concern me, for now (I’ll put aside the further complexity of hallucinations  for a later post).

Thus, the possibility concerns only the properties that I seem to perceive. With that, then, we say this:

It is possible that every property which I seem perceive or seem to have ever perceived is merely apparent. There is no and never was such a property (at least, properly connected to me via perception and memory — but take that as implied in what I say hereon). Let us call this possibility the possibility of Universal Illusion: ‘illusion’ because there is the appearance of a property without there being a real property; ‘universal’ because it applies to all properties which one seems to experience or to have experienced.

The question then is this: Does one have to accommodate the possibility of Universal Illusion in one’s theory of perception?

We might ask the question this way: When one explains how one’s perception works, including how certain appearances arise, and how they connect to the external world, must one include an account which allows that any appearance of a property can occur without there being such a property?

Or we might ask it this way: In providing an analysis of perception as a set of features,  including phenomenology, causation, and conditions for how it actually occurs,  must one include in that analysis, have it built into the system, the possibility to generate universal illusion — even if, as one assumes of one’s actual perception, no such universal illusion ever occurs?

In answer to all of these, I would say: we need not cover the possibility of universal illusion in our perceptual theories.

Actual capacity

Something, such as universal illusion, may be possible, but if it is only merely possible, one does not have to include in what is actual. How our perception — yours, mine — works is something actual; the range of appearances it can give rise to is an actual range, not a merely possible range. The merely possible is useful when developing accounts of this range. But nothing merely possible is necessary for it. So, for any appearance of x, it may be possible that there is no real x. But, if in providing a theory of perception, I do not accommodate that possibility, with no other motivation to do so (see below) then this is not a failure of my theory. If one presents a perceptual system which does not give rise to universal illusion, one does not fail to provide a suitable perceptual system for our actual perception.

Why would you chose a perceptual system which does not have the capacity for any illusion over one that does? I will discuss this in detail in another post (on the importance of appearances for empirical theory). But briefly, here, my view is this:

  • The positing of illusion is either irrelevant to a theory, where the theory has no empirical content, or is a bad-making feature of a theory with empirical content: the more one’s theory relegates appearances to illusions, the worse it is as an empirical theory. This is to do with the role appearances play in an empirical theory.
  • So, perhaps we have a choice between (a) a perceptual theory which gives rise to any illusion or (b) a perceptual theory which restricts what kinds of illusion can arise. If one is free to chose, for the empirical reasons above, one should select ‘b’: it limits the number of illusions by limiting the number of illusions that can arise; one can say, given (b), in certain cases of appearance, there is no illusion here because there is never illusion here.

More is to be said on this. Here, I assume that a theory of perception need not provide an account involving the capacity for universal illusion, and, even more, the less illusions it is committed to the better. If one thinks we need perception to have the capacity for universal illusion, one needs more than the mere possibility of universal illusion. One needs something to push us in this direction.

Actual illusions

One would fail to provide a good perceptual theory if one did not account for actual cases of illusion. In such a case, there are actual cases of perceptual illusion which are not accounted for by your perceptual theory. Such actual perceptual illusions are cases of a certain relationship between perceptual experience and the world otherwise (in the case of illusion, the relationship is at least a discrepancy between perceptual appearance and the world).

This leads to my next assumption about theories of perception: perceptual theories need not be about perception in a void; they need only be theories about perception in the world (they can, of course, be consistent with a theory of perception in a void). This means, first, a perceptual theory which does not account for the actual range of relationships is a perceptual theory fails to account for something actual. If one element in that range is an illusion, the theory fails to account for an actual illusion.

But it also means that, should a theory require that, in some cases, perception is constituted by worldly entities, things which are impossible given perception in a void, this is not a weakness of the theory. I assume that it is not a condition of a perceptual theory that it be consistent with Descartes’ imagined scenario of an evil demon, nor with a solipsist’s “world”(if it can be called that).

This suggests the following: one’s perceptual theory need only be worldly; further, they have an empirical aspect to them concerning actual empirical evidence. It is tested by actual illusions, not merely possible ones.

This is not exhaustive of what perceptual theories have to explain. But it is, I think, an important part of what must be uncontroversially explained. A reductive or eliminativist physicalist might need to reduce or eliminate  ‘perception’ itself and, in so doing, explain why we refer to it. But this is something which other theories will not be concerned with.


1. I might also take back what I said and explain that in fact there is no illusion here. There is no discrepancy between how things seem and how things are. There is a discrepancy between how things are and something other than how things seem. This ‘other’ is what makes me believe or want to report that I see K, P, R, x, a combination of these, etc.